The possibility of using visible and short-wavelength near-infrared (SW-NIR: 600 to 1100 nm) spectroscopy to detect the onset of spoilage and to quantify microbial loads in rainbow trout (Oncorhynchus mykiss) was investigated. Spectra were acquired on the skin and flesh side of intact trout fillet portions and on minced trout muscle samples stored at 4 °C for up to 8 d or at room temperature (21 °C) for 24 h. Principal component analysis (PCA) and partial least squares (PLS) chemometric models were developed to predict the onset and degree of spoilage. PCA results showed clear segregation between the control (day 1) and the samples held 4 d or longer at 4 °C. Clear segregation was observed for samples stored 10 h or longer at 21 °C compared with the control (0 h), indicating that onset of spoilage could be detected with this method. Quantitative PLS prediction models for microbial loads were established. For trout fillets, 4 °C: R= 0.97, standard error of prediction (SEP) = 0.38 log colony-forming units (CFU)/g (flesh side); R= 0.94, SEP = 0.53 log CFU/g (skin side); R= 0.82, SEP = 0.82 log CFU/g for minced fish held at 21 °C. These results indicate that SW-NIR in combination with multivariate statistical methods can be used to detect and monitor the spoilage process in rainbow trout and quantify microbial loads rapidly and accurately.